PC processors have bitwise instructions processing floats.

When I need absolute value, I usually do it like that in C++:

    return _mm_andnot_pd( _mm_set1_pd( -0.0 ), vec );